> -----Original Message----- > From: Kai MacTane [mailto:[EMAIL PROTECTED] > Sent: 25 July 2003 17:34 > To: [EMAIL PROTECTED] > Subject: RE: [SAtalk] Oops... running 2.60 > > > Actually, I doubt those BAYES_00 hits are doing you much > good, either. If > messages like these trigger BAYES_00, I have to wonder if > there's something > wrong with your Bayes tokens/db/etc. > I wipe the bayes db every couple of weeks to avoid this (over time it starts giving more and more FNs). I wiped it again just after sending the message, so it'll take a little while before the BAYES_00 creeps back again.
I went to 2.60 because there was more spam getting through than getting caught
with 2.55. 2.60 is still about 75% effective, so it's still useful. Those HTML only spams
are a complete git though (I had >50 of them in my inbox this morning... Ended up deleting the
entire inbox and emailing people to resend if they had something I needed to see, which is what I used to
have to do before SA existed).
So you're saying that 2.55 was catching less than 50% of your spam ?
Honestly, you must have something wrong with your setup causing that, theres no way that SA's hit rate is that poor. Here we're getting at least 95% detected, and thats with a threshold of 7.0 instead of the default 5.0
If you're wiping the bayes database every couple of weeks then that wont be helping either. You need to find out why it is giving you FN's perhaps you don't have suitable autolearn thresholds.
Try using check_bayes_db and browse through the token database to see if anything strange is being learnt, and/or run some troublesome spams through spamassassin in debug mode to see what tokens are being scored, and in which direction...
Another thing you could try is turn autolearning off and learn the biggest bunch of spam and ham you can and see how you go.
If you're wiping your bayes database all the time and then letting it learn again, then until it has autolearnt 200 hams and 200 spams you're not using bayes, and in my experience SA with bayes disabled is MUCH worse.
Also, are you using RBL checks and DCC, Razor, Pyzor ?
If you're not, that could easily explain your poor hit rate especially with HTML only spams - to catch those effectively you really need the RBL checks and DCC, Razor, and Pyzor all working...
Regards, Simon
------------------------------------------------------- This SF.Net email sponsored by: Free pre-built ASP.NET sites including Data Reports, E-commerce, Portals, and Forums are available now. Download today and enter to win an XBOX or Visual Studio .NET. http://aspnet.click-url.com/go/psa00100003ave/direct;at.aspnet_072303_01/01 _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
